INTELLIGENCE,
DEPLOYED FROM WITHIN.

A boutique team of forward-deployed engineers building bespoke superintelligence systems — for organizations of any size that demand frontier capability without the frontier headcount.

Scroll
// SUPERINTELLIGENCE.AS.A.SERVICE
Superintelligence-grade systems. Engineered from within.
// WHAT.WE.BUILD

Cost-aware multi-agent systems that research, decide, and act at machine speed. Compounding intelligence that sharpens with every cycle. Provenance-grounded knowledge graphs — a company brain your systems can reason over.

// WHO.WE.BUILD.FOR

Organizations that require more than a vendor. From institutional leaders in quantitative finance, energy, and healthcare to ambitious teams of any size — those who demand frontier capability and own it.

// AGENTIC.ENGINEERING

Autonomous systems that research, decide, and act.

We deploy cost-efficient multi-agent systems that close the gap between research and execution. They ingest live data, reason over your context graph, and act within defined parameters — a continuous, self-improving extension of your team.

// THE.FIRM

Elite Quants.
Agentic Engineers.
Embedded Partners.

Founded in Montecito, California in 2008. Obsidian Quantitative is a boutique firm where elite quantitative research meets agentic SI engineering. We operate as forward-deployed engineers, embedding directly into client operations to deliver superintelligence-grade systems. Engagements are limited and by fit.

// ESTABLISHED
2008
// BASED.IN
Montecito, California
// PRINCIPALS
2
// THE.PRINCIPALS
Alex Bernal
Founder · Principal SI Engineer & Quant

Sets firm strategy and research architecture, and designs the agentic systems we deploy. Twenty years across quantitative finance, data science, and frontier AI — now forward-deployed, building cost-aware multi-agent systems and compounding-intelligence architectures inside client operations. MS Financial Engineering · MS Applied Data Science · CMT · CFTe.

Ben Balaj
Lead SI & Data Engineer

Builds and deploys the firm’s agent harnesses, data pipelines, and context-graph infrastructure — the systems that make autonomous agents reliable in production. Specializes in model routing, caching, and eval harnesses for cost-efficient agentic deployment. BS Finance · Bloomberg Certified · Alteryx · Python.

// WHAT.WE.DO
01
AGENTIC SI ENGINEERING
Cost-aware multi-agent systems and full-company agentic deployments — any role, any scale. Orchestrator-worker architectures built across the full agentic stack: Anthropic, OpenAI, Grok, Hermes, and custom orchestration.
02
COMPOUNDING INTELLIGENCE
Self-improving systems that sharpen with every cycle, grounded on an auditable, provenance-grounded context graph — a company brain your agents reason over. Digital-twin simulation to design and tune before production.
03
QUANTITATIVE RESEARCH & EMBEDDED DELIVERY
Elite factor models, systematic strategies, and signal engineering — delivered forward-deployed. We embed in your stack, from AWS and Azure to specialized infrastructure, and run as part of your team.
// ENGINEERING.EDGE
—
HARNESS ENGINEERING
Custom agent harnesses, eval rigs, and tooling that make autonomous systems reliable in production.
—
DATA SOVEREIGNTY
Proprietary data preserved, isolated, and never surrendered to third-party models.
—
MODEL ROUTING
The right model for each task, dynamically routed across frontier providers.
—
ELITE CACHING
Context and prompt caching that cuts agentic cost by orders of magnitude.
// SECTORS.SERVED

Financial Markets  ·  Healthcare  ·  Energy  ·  Private Equity  ·  Capital Markets

Initiate an Engagement →
// TECHNOLOGY.STACK

Core capabilities.

// PRIMARY.FRONTIER.MODELS
ANTHROPIC
  • Claude Code
  • Claude for Work
  • Agentic Pipeline Engineering
  • MCP + Tool Use
  • Constitutional AI Systems
OPENAI
  • GPT-5 class models
  • Agents SDK
  • Realtime API
  • Function Calling
  • Fine-tuning
GOOGLE
  • Gemini frontier models
  • Vertex AI
  • BigQuery ML
  • TPU Infrastructure
  • Google Cloud AI
xAI
  • Grok frontier models
  • Real-time X data
  • Grok API
  • Reasoning models
NOUS / HERMES
  • Hermes open-weight models
  • Sovereign / on-prem deploy
  • Fine-tuned agents
  • Data-sovereign reasoning
EVALS & OBSERVABILITY
  • LLM-as-judge rubrics
  • Agent tracing & telemetry
  • Custom eval harnesses
  • Regression & drift detection
  • LangSmith · Braintrust
// AGENT.STACK

MCP (Model Context Protocol)  ·  LangGraph  ·  OpenAI Agents SDK  ·  Multi-agent orchestration  ·  GraphRAG context graphs  ·  Eval & observability harnesses  ·  Model routing & caching

// INFRASTRUCTURE
NVIDIA
  • CUDA-X AI
  • NVIDIA NIM
  • TensorRT-LLM
  • AI Enterprise
  • Omniverse
MICROSOFT AZURE
  • Azure OpenAI
  • Copilot Studio
  • Azure ML
  • Cognitive Services
  • Q Business
AMAZON WEB SERVICES
  • AWS Bedrock
  • SageMaker
  • Rekognition
  • Comprehend
  • Q Business
// HOW.WE.ENGAGE
Selective. Integrated. Institutional.

Founded in Montecito, California, Obsidian Quantitative operates as a high-conviction research and engineering firm. Every engagement is structured as an integrated partnership. We provide the elite quantitative research and embedded SI engineering required to maintain a competitive edge in complex markets.

// SELECT.ENGAGEMENTS

Limited institutional partnerships across financial markets, healthcare,
energy, and private equity.
Client details available under NDA.

// INITIATE.CONTACT

Initiate an engagement.

Engagements are limited and by fit. Qualified inquiries welcome.

Inquiry received.